"globalIllumination": simulates global illumination of an apartment scene using simplified photon mapping. Its output geometry and lightmaps can be used by FlatMatch to created a globally illuminated 3D scene of the apartment to explore interactively.

Dependencies:

• gcc (C and C++ frontends)
• GNU make
• headers and libraries for libpng, e.g. the Ubuntu package libpng-dev
• OpenCL development headers, e.g. the Ubuntu package opencl-headers
• either
  • an OpenCL implementation ("ICD") of your GPU vendor (AMD and Nvidia have been tested), or
  • an OpenCL implementation supporting your CPU (AMD and Intel SDKs have been tested)
    • e.g. the Ubuntu package beignet-dev

Build instructions:

• to select a CPU-based implementation change both occurrences of CL_DEVICE_TYPE_GPU in global_illumination_cl.c to CL_DEVICE_TYPE_CPU
• make sure that only one OpenCL ICD is installed (otherwise the code around clGetPlatformID would need to be modified)
• "make" (for GPU-based computations)
• Note: this project mixes C and C++ code. This is not a good coding style, but happened when adding the C API-based OpenCL renderer

Compilation Output:

• executable globalIllumination

Usage:

• ./globalIllumination <layout image> reads the color-coded apartment layout (see parseLayout.c for the color hex codes used to color-code the apartment geometry) PNG image from , and computes apartment geometry and lightmaps
• alternatively, call ./generate_flatmatch_entry.py <layout_image> <offer_id> <scale> <latitude> <longitude> <yaw> <level> with
  • <layout_image> the path to the source file in PNG format, for example the included example.png
  • <offer_id> an arbitrary numerical id
  • <scale> the scale of the layout image, in pixels/m
  • <latitude> the apartment's geographic latitude
  • <longitude> the apartment's geographic longitude
  • <yaw> the geographic direction (in degrees) in which "up" in the layout image is pointing (0 is north, 90 is east, etc.).
  • <level> the building floor/level at which the apartment is located. Zero is the ground floor, a level of "x" moves the apartment floor 3.5*x meters above the ground

Output:

• the first call generates
  • set of lightmap textures in tiles/
  • an RLE-encoded collision map collisionMap.json of the same dimensions as the input image, stored as JSON. Each number in the JSON array determines how many successive pixels are passable/impassable. The map starts with passable pixels and then alternates between impassable and passable.
  • the 3D apartment geometry in geometry.json. Each entry in the "geometry" array describes a rectangle in 3D space. Texture coordinates are implicit, i.e. for each rectangle the point
    • "pos" has texture coordinates (0,0)
    • "pos + width" has texture coordinates (1, 0)
    • "pos + height" has texture coordinates (0, 1)
    • "pos + width + height" has texture coordinates (1, 1)
• the second call generates a full REST folder structure that can directly be used in a FlatMatch web application via the URL flatview.html?rowid=<offer_id>